Previous | Next --- Slide 24 of 54
Back to Lecture Thumbnails

I would guess that internet traffic as a whole tends to increase around Christmas. Does this simply push the load of having surplus servers one step back from most companies to AWS and like providers? If Amazon needs more servers at a given time, it has less to offer on AWS.


Well most companies use AWS exclusively, as the cost of aws is usually smaller than having a data center, servers, and an it dedicated to just the website.


So, to achieve elasticity we need a source of servers available on-demand, what about the server providers themselves? Does it mean that cloud services like AWS, Google Cloud and MicroSoft Azure also have this burst patterns and will need to keep the number of servers much higher than average and have most of them wasted on normal occasions?


I agree with @jerryzh. I would guess that at some point some services are not able to utilize elasticity effectively since if everyone was using the concept of elasticity, there needs to be someone who can provide to all that change in demand.


The little I know: The idea is to make clients of different traffic patterns share the same physical resource so that when there is a traffic burst for one client the other has minimal traffic and therefore doesn't really need as much resources.


AWS has a service called Auto Scaling, which can be used to scale EC2 instances up or down automatically according to some scaling policies like adding a new instance when the average CPU usage is high. More information can be found here.


@jerryzh: Were you able to figure out the answer to your above question? If yes, please post! Thanks.


@randomized, I think @Chrome's comment in slide 30 may help.